Why Synthetic Users, and Why Now?

Here's a sobering reality for market researchers: While your business has learned to pivot marketing strategies weekly, launch products in months instead of years, and respond to trending topics in real-time—your research timelines have barely improved.

Recruiting participants still takes weeks. Fielding and analyzing studies can drag on for over a month. By the time insights reach decision-makers, the market has already shifted.

A Qualtrics survey of 3,000+ researchers found that 62% of organizations depend on research insights significantly more than they did two years ago.

Yet teams face rising pressure to deliver those insights faster without sacrificing quality.

That’s where synthetic users come in. The question isn’t why we need them—it’s why now.

The Perfect Storm: Three Forces Converging

The emergence of synthetic users isn't random. It's the result of three powerful forces converging at exactly the right moment:

1. Business Pressure: The Need for Speed

Modern businesses operate at a pace that traditional research simply can't match.

Consider the CPG industry: Half of the top CPG companies reduced headcount or froze hiring in 2023, while EBIT margins hit 10-year lows at 12.2%. Yet the same executives identified developing "a robust insights engine" as critical to competitive advantage.

The math doesn't add up: fewer resources + growing dependence on insights + faster market cycles = an impossible equation with traditional methods.

In a world where:

• Marketing messages change weekly based on trending topics
• Product launches happen in quarterly sprints, not annual cycles
• Competitors can copy your innovation within months

...waiting six weeks for concept test results isn't just slow—it's strategically untenable.

💡 The Speed Gap

Traditional research operates on a timeline of weeks to months.
Modern business decisions happen on a timeline of days to weeks.

This gap isn't narrowing—it's widening. And that's creating existential pressure for the research industry.

2. AI Progress: LLMs Can Now Simulate Nuanced Behavior

Five years ago, AI-generated text was clunky and unconvincing. Three years ago, it was impressive but still noticeably artificial. Today? Large language models like GPT-4, Claude, and Gemini can generate responses that are often indistinguishable from human writing.

But more importantly for market research, they can now:

Understand context and nuance: They don't just match keywords—they grasp the difference between "I might buy this" and "I'd definitely buy this if it were on sale."

Model demographic and psychographic personas: They can simulate how a 45-year-old suburban parent who shops at Whole Foods would respond differently than a 28-year-old urban professional who orders from Instacart.

Maintain consistency across conversations: They can answer follow-up questions without contradicting their initial responses, behaving like a coherent persona rather than a random text generator.

A breakthrough study from October 2024 demonstrated that LLM-based synthetic users achieved 90% of human test-retest reliability when tested on 57 consumer product surveys. That's not perfect replication—but it's close enough to be genuinely useful for early-stage testing and iteration.

Meanwhile, Harvard Business School researchers showed that GPT-derived willingness-to-pay estimates are realistic and comparable to human studies—especially when fine-tuned with previous survey data.

The AI hasn't just gotten better. It's crossed a threshold where it's reliable enough for practical use.

3. Data Abundance: The Fuel That Powers Simulation

Synthetic users don't work in a vacuum. They need data to learn from—and we're living in an unprecedented era of behavioral data availability.

Consider what's now accessible:

• Decades of consumer survey data from major research firms
• Millions of product reviews across e-commerce platforms
• Behavioral tracking data from loyalty programs and apps
• Social media conversations revealing real-world attitudes and preferences
• Academic datasets from psychology, marketing, and consumer behavior research

In September 2024, NielsenIQ launched BASES AI, a platform that uses "consumer-permissioned data on consumption patterns across hundreds of thousands of households" to create synthetic respondents. This isn't data scraped from the internet—it's real, structured behavioral data from actual purchase decisions.

When companies like NIQ, with their vast repositories of consumer data, invest in synthetic user technology, it signals that the data infrastructure is finally mature enough to power realistic simulations.

[Diagram placeholder: Three pillars—Business Pressure, AI Maturity, Data Availability—converging to create the "Synthetic Users Moment"]

The Macro Trends Accelerating Adoption

Beyond these three foundational forces, several broader trends are pushing synthetic users from experimental to essential:

Democratization of insights: Not just research teams, but product managers, marketers, and even sales teams want access to consumer feedback. Synthetic users make insights accessible to more stakeholders without overwhelming research departments.

Generative AI normalization: As ChatGPT, Midjourney, and other gen-AI tools become everyday business tools, the conceptual leap to "AI-powered consumer simulations" feels less radical and more inevitable.

Shortened innovation cycles: CPG companies that once launched 3-4 new products per year are now testing 10+ concepts per quarter. Traditional research can't keep pace with this volume.

Economic pressure: With marketing budgets under scrutiny and research seen as a cost center rather than revenue driver, the promise of "10x faster insights at 1/5th the cost" is impossible to ignore.

What This Looks Like in Practice

Let's make this concrete. Imagine a CPG company preparing for their quarterly innovation review. Traditionally, they might:

• Develop 10 product concepts
• Research budget allows testing 2-3 concepts
• Wait 6 weeks for results
• Pick 1 winner to move forward
• Hope they chose correctly

With synthetic users, the same company can:

• Develop 10 product concepts
• Test all 10 with 500 synthetic consumers overnight
• Identify the top 3 performers by Tuesday morning
• Validate those 3 with a targeted human study (faster and cheaper since the field is narrowed)
• Move forward with confidence that they tested comprehensively

A real example: When EY tested Evidenza's synthetic consumer technology against their actual annual brand survey of senior executives, they found 95% correlation between synthetic and real responses—delivered in days rather than months, at a fraction of the cost.

Companies like Mars, Salesforce, and EY aren't dabbling with synthetic users as a curiosity. They're integrating them into core research workflows because the business case is undeniable.

The Inflection Point

We're at a rare moment in the evolution of market research—similar to when online surveys disrupted mail panels in the early 2000s, or when mobile research transformed data collection in the 2010s.

The difference? This shift is happening faster because the underlying technology (LLMs) is improving at an exponential rate, and the business pressure (speed, cost, agility) has never been more intense.

💡 Why This Moment Matters

Synthetic users aren't the future of research—they're the present. The question isn't whether they'll become mainstream, but how quickly, and whether your organization will be an early adopter or a late follower.

📌 Key Takeaways

✅ Business demands have outpaced research capabilities, creating an unsustainable gap between insight needs and delivery timelines

✅ AI has matured to the point where LLMs can simulate realistic consumer behavior with 85-90% accuracy vs. human responses

✅ Rich behavioral datasets now provide the fuel needed to train credible synthetic consumers

✅ Major brands are already adopting synthetic users (Mars, EY, Salesforce), signaling this isn't experimental—it's practical

✅ We're at an inflection point where synthetic users are transitioning from novel to necessary

➡️ What's Next?

You now understand What are synthetic users and why they're emerging now. But what specific problems do they solve for market researchers like you?

In Chapter 3: What Problems Do Synthetic Users Solve?, we'll dive into the concrete pain points of traditional research—and show exactly how synthetic users address each one.